Technology development and deployment decisions are justified by weighing their costs against the expected benefits. Multiple nuclear fuel cycle (NFC) simulation models have been devised, some with the aim of quantifying cyclewide sensitivities to variations from base-case scenarios. Base-case sensitivity studies often perturb only one parameter at a time and only in the region around the initial value. This paper details a sensitivity study methodology that applies entropy-based statistical methods of information theory to describe outcomes produced by an NFC model. This supersedes past efforts at sensitivity and uncertainty analysis by allowing a much larger space to be explored. Here, 30 independent fuel cycle parameters for a fast reactor light water reactor hybrid scenario are varied simultaneously and stochastically. This fuel cycle schema was chosen as a well-known, sufficiently complex model; the underlying statistical methods could be applied to any cycle. This study uses the uncertainty coefficient computed from contingency tables (CTs) to represent the sensitivity of a technology-defining input to the response. The response of interest here was taken to be the deep geologic repository capacity for a given realization of fuel cycle inputs. After computing the uncertainty coefficients, the inputs themselves are sorted based on decreasing sensitivities. Fast reactor used fuel plutonium separations were found to be most important to the cycle. Furthermore, to represent input covariances (the effect of one input on the sensitivity of a second input to the response), a new measure is defined on three-dimensional CTs. This metric is the coefficient of the variation of uncertainty coefficient of two-dimensional slices of the original table. Sorting by this sensitivity of sensitivity metric, the input pair of fast reactor americium separations together with high-level-waste storage time was found to have the largest joint effect on the repository capacity.